|
|
|
|
|
|
|
|
|
|
xen-users
Re: [Xen-users] Checkpointing
tbrown@xxxxxxxxxxxxx wrote:
> network state is part of the kernel memory that is saved and/or
> migrated.... so that doesn't seem to be an issue.
Right but that is only network state as of when the kernel memory was
saved. The tcp stream (for example) has moved on since then. If I am
scp'ing a file and suddenly the server reverts to the state of 5 seconds
ago, the connection will die for sure. If the incorrect TCP sequence
numbers didn't kill it the cipher block chaining of ssh certainly would.
You basically have to constantly stream the current network state to
some storage medium in real time. Around 6 years ago I had a pair of
Network Appliance machines...F750 I think they were. They would actually
do this. They came with some big fat cable you would connect them with
and they would maintain the state of every network transaction in real
time over that interconnect so that if one machine died the other could
take over without dropping the connections. It was pretty sweet. Perhaps
some day xen will be able to achieve this sort of high availability
using a similar technique.
--
Tracy R Reed
http://ultraviolet.org
_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-users
|
<Prev in Thread] |
Current Thread |
[Next in Thread> |
- Re: [Xen-users] Checkpointing,
Tracy R Reed <=
|
|
|
|
|